Data Comparisons

Death Rate in Illinois

18 Comments

Many people here are discussiong the death rate for the disease. I plotted data from https://raw.githubusercontent.com/COVID19Tracking/covid-tracking-data/master/data/states_daily_4pm_et.csv which contains data for “new cases” and “new deaths”. I pulled data for Illinois, then plotted the ratio over time. Numerator and denominator are smoothed before plotting.

Here’s what I see: Right now, in Illinois, in a periods where case rates and death rates seem to be growing slowly (rather than explosively), the ratio of “deaths”/”detected cases” appears to be fall in the ballpark of 3%-5%.

I know people are going to ask:

  1. Can I provide separate traces for age groups? Nope.
  2. Can I provide separate traces for ethnic or racial groups? Nope.
  3. Can I correct for undetected utterly symptomless cases? Nope.
  4. Can I correct for some other “RBI against right handed pitcher during the 9th inning with bases loaded” type issue? Nope.

The reason I can’t is that, at least for the time being, I don’t have access to the appropriate data. If you know where those data are, let me know. (Not: please be somewhat specific. Don’t just give me a link to the top of a big page with lots of displays or fling out the name of a researcher. If you haven’t found a link to the underlying data, that’s not data.)

18 thoughts on “Death Rate in Illinois”

  1. Here is something curious about Case Fatality Rate (CFR): It correlates reasonably well with the logarithm of Deaths per 1000 population (DpK).

    CFR = (7.3 ± 1.4 %) + (1.26 ± 0.15) * ln(DpK)

    R = 0.66
    R^2 = 0.44

    Error bars are standard errors including covariance. I used cumulative state level data from Real Clear Politics through May 13.
    .
    I don’t know what it means.

  2. Mike M,
    I have no idea what your equation is supposed to mean. I know it’s not data.

    Can you explain what that is? Is it a fit? to some data? If so, which? Is it some theoretical model based on… whatever?

    Is deaths per 1000 cummulative? Daily? Whose data did you use for CFR and so on? (Obviously, we haven’t hit the final numbers for cummulative death rates!)

    Real questions.

  4. Lucia,
    I am a little puzzled by your graph. There is a substantial offset in time between confirming a case and the patient’s death. Should you not be forming a ratio between recent deaths and new cases from 7 to 10 days earlier? Unless the time offset is taken into account, the ratio of deaths to cases will be driven mainly by raising and falling number of new cases, and may not tell you much about the true ratio. Early in the epidemic, the cases are rising much faster than the deaths, so the ratio is low. Once you pass the peak in cases, the deaths keep rising while the number of new cases starts falling, which gives you too high a ratio. In the declining part of the curve, the ratio will always a little higher than correct, because deaths are always lagging behind the new cases.

  5. SteveF,
    Yes. There should be an offset. I don’t know what that offset should be exactly. If we were at a “steady state” situation, you wouldn’t expect the offset to matter. (Well… except in the current case, there is the issue of whether our ability to detect cases changes too.)

    Early in the epidemic, the cases are rising much faster than the deaths, so the ratio is low.

    Yes. that’s my interpretation of what’s happening early on.
    .
    I think the main thing this shows is order of magnitude. But no, it doesn’t show case mortality without having some issues. But still… order of magnitude.

  7. Lucia,
    I grabbed those numbers. If all you are looking for is a good order of magnitude value, you can just divide the cumulative deaths until the date you are plotting by the cumulative confirmed cases until 7 days before that date…. It is not far from 5%, and has been since very early in the pandemic (for Illinois). There are caveats, of course.
    .
    But even if that value represents everyone who died relative to everyone who felt sick enough to be tested, I’m not sure why you want to know that number. The number I think would be informative is cumulative deaths until 10 days ago divided by total seropositives in the entire population today, since that is probably the best estimate of true IFR. NY state probably has by now 5 million seropositives among the adult population and had about 25,000 deaths a week ago, so an inferred IFR (among adults) of about 0.5%. add in children, and the rate for the whole population drops to about 0.35%. There are caveats here as well, of course.
    .
    The drop in rates in NY has been pretty dramatic compared to states with far lower death rates, and since I doubt social distancing is much more effective in NY than elsewhere, it seems to me very likely that the NY death rate has fallen in large part because herd immunity in the NYC area is contributing substantially to the drop in Rt. As NY residents become less locked down, we can expect the rate of deaths to drop more slowly (or even rise), but if herd immunity is important, there should be proportionally more deaths outside the NYC area going forward.

  8. A paper in Cell suggested in 2016 that there are significant differences in immune response between European and African populations, stemming in part from admixture of Neanderthal genes in Europe. The Africans have a stronger response to viral infections (like influenza), but a greater susceptibility to autoimmune disease like lupus. Does that help explain low infection rates in Africa? Does that predispose people of African descent to ‘cytokine storms’ as well? Dono.
    DOI : 10.1016/j.cell.2016.09.024

  10. SteveF,
    Yeah, let’s bring up racial genetic superiority, that always goes over well, ha ha. Seriously though this could provide important clues to any genetic factors. However anyone in academia without bullet proof tenure won’t touch this with a ten foot pole. African Americans are dying at higher rates per capita in the US and they don’t really understand why. Some of it is higher prevalence of underlying conditions, less access to healthcare, etc. Even bringing up the role of genetics will get one instantly excommunicated out of polite society without a trial. Academia and the media can set whatever standards they like, but the “do not even look there” attitude isn’t helpful as a science standard.
    .
    As for the socially acceptable media takes on the situation, here is one of the most popular articles from The Atlantic last week:
    The Coronavirus Was an Emergency Until Trump Found Out Who Was Dying
    The pandemic has exposed the bitter terms of our racial contract, which deems certain lives of greater value than others.
    https://www.theatlantic.com/ideas/archive/2020/05/americas-racial-contract-showing/611389/
    .
    This viewpoint from left activists is so common now that it is quite boring. The idea that everything is to be viewed through the original sin of race/slavery won the NYT a Pulitzer Prize (The 1619 Project). It’s very divisive and unhelpful IMO.

  11. SteveF

    by total seropositives in the entire population today,

    The problem is we don’t have that most places. I don’t think I ever read details of the NY study. Do we know the false positive rate of the test given? (I don’t– but does someone?)
    I agree I’d rather have that number. But unless I really understand the provenance of the numbers used to go from a case death rate to an infective death rate, I don’t trust the calculations.

    The one in the figures is not of greatest interest. But at least I trust the numerators and denominators more.

  12. Lucia,
    As far as I know no details of the NY antibody study (like false positive rate) were ever published. Maybe they were and it was not well publicized, or maybe the details will be released later.
    .
    I do know that the study showed ~22% positive prevalence in the NYC region (I don’t know what that region included), but only 3.5% outside the city region. That puts an upper bond on false positives of 3.5%, and probably much lower than that, since there have definitely been lots of cases outside the city region. Without details of the study, it is impossible to say more than that.
    .
    It does seem to me there was a rapid loss in interest in antibody screening studies after the handful that were done. I could speculate on why, but I will leave it at that. I will be very surprised if the ever growing fraction of the population with resistance has not been responsible for at least a substantial part of the drop in new cases and deaths in NY.

  13. Tom Scharf,
    “Even bringing up the role of genetics will get one instantly excommunicated out of polite society without a trial.”
    .
    For someone who wants a rational understanding of reality, the only sensible reply to that is a question: “What Do You Care What Other People Think?” (with credit to R. Feynman, of course). The nutcakes on the left who shout “racist!” endlessly, are a lot worse than boring, they are a danger to both science and rational knowledge in general. I really could not care less what such fools think.

  14. SteveF

    It does seem to me there was a rapid loss in interest in antibody screening studies after the handful that were done.

    I think this is a shame. Good serologic studies would give us information very much worth knowing!

  15. From an article on vaccine development:
    https://blogs.sciencemag.org/pipeline/archives/2020/05/15/good-news-on-the-human-immune-response-to-the-coronavirus

    in the unexposed patients, 40 to 60% had CD4+ cells that already respond to the new coronavirus. This doesn’t mean that people have already been exposed to it per se, of course – immune crossreactivity is very much a thing, and it would appear that many people have already raised a response to other antigens that could be partially protective against this new virus.

    CD4+ cells, a.k.a. T helper cells, are a critical part of the adaptive immune system. I don’t know if it possible that people with CD4+ cells that respond to the Wuhan virus might actually be immune rather than partially protective.

  16. Mike M,
    There have already been a couple of studies about the apparent pre-existing cross-reactivity, including discussions on earlier threads of this blog. There is a very good chance previous exposure to other coronaviruses infers some resistance, if not protection. Unfortunately, it will likely be a year or two before this subject is figured out. The fall-off of cases in NYC long before all potential victims were infected, suggests to me there is significant pre-existing resistance.

  17. SteveF,

    I thought the earlier discussions were about antibodies, were not as specific, and were not as high as 40-60% of the population. But my memory is not what it once was.
    .
    The virus need not infect all, or even most, potential victims for the rate of new infections to fall off. It depends on how contagious the virus is. So far as I can tell, the Wuhan virus is not all that contagious.
    .
    Yes, I know the claim that it is extremely contagious has been repeated a zillion times. But repetition is not evidence. I have not seen any evidence in support of that claim, other than the rapid initial spread. I suspect that is evidence of the vulnerability of an unexposed population, not the infectivity of the virus.
    .
    Note that the latter point has been indirectly raised a couple of times here. First when I called attention to the paper by Gomes et al., then when someone else called attention to a Climate Etc. post by Nic Lewis on the same subject.

  18. More on helper T-cell cross reactivity from the WSJ in a piece about the latest scary media meme seemingly designed to support permanent lockdowns:

    The Evidence on Kids and Covid
    Rare serious cases bear watching, but overall risk appears to be low.

    It’s also possible that children have cross-protection from earlier coronavirus-causing common colds. Two new studies find that antibodies and white blood cells that individuals developed to fight recent cold infections also react against the novel coronavirus. La Jolla Institute for Immunology researchers discovered that about half of stored blood samples from 2015 to 2018 carried helper T-cells that react against the novel coronavirus.

    Yet unlike with other respiratory viruses, children don’t appear to be large spreaders. Australia’s National Centre for Immunisation Research and Surveillance tracked Covid-19 cases at 15 schools from March 1 to April 16. At the outset, 18 individuals were infected. After six weeks only two of their 863 close contacts at the schools had become infected. There were no cases of students passing the virus to teachers. Studies from China show kids were more likely to pick up the virus from their parents than vice versa. While Taiwan has lower case and fatality rates than Western countries, it did not impose widespread school closures.

    Edit: Oops. This isn’t really more on helper T-cells. It’s based on the same study linked above.

Comments are closed.